Clinical Clinical Computed Tomography Research Articles

Automated segmentation of clinical CT scans of the cochlea and analysis of the cochlea's vertical profile

PurposeKnowledge of the cochlear anatomy in individual patients is helpful for improving electrode selection and placement during cochlear implantation, as well as in surgical planning. The aim of this study was to develop a model-free automated segmentation algorithm to obtain 3D surfaces from clinical computed tomography (CT) scans that describe an individual's cochlear anatomy and can be used to quantitatively analyze the cochlea's vertical trajectory. MethodsClinical CT scans were re-oriented and re-sliced to obtain mid-modiolar slices. Using these slices, we segmented the cross-section of the cochlea. Results3D surfaces were obtained for the first 1.5 turns of 648 cochleae. Validation of our algorithm against the manually segmented ground truth obtained from 8 micro-CT scans showed good agreement, with 90 % area overlap and an average distance of 0.11 mm between the segmentation contours. The average cochlear duct length for the basal turn was 16.1 mm along the central path and 22.4 mm along the outer wall. The use of an intrinsic, observer-independent coordinate system and principal component analysis enabled unambiguous quantitative evaluation of the vertical trajectory of the cochlea, revealing only a weak correlation between the symmetry of the commonly used basal turn diameters (B-ratio of A and B diameters) and the profile of the vertical trajectory. ConclusionA model-free segmentation algorithm can achieve similar accuracy as previously published methods relying on statistical shapes. Quantitative analysis of the vertical trajectory can replace the categorization into rollercoaster, sloping, or intermediate vertical trajectory types.

Deep Learning Method for Rapid Simultaneous Multistructure Temporal Bone Segmentation.

To develop and validate a deep learning algorithm for the automated segmentation of key temporal bone structures from clinical computed tomography (CT) data sets. Cross-sectional study. A total of 325 CT scans from a clinical database. A state-of-the-art deep learning (DL) algorithm (SwinUNETR) was used to train a prediction model for rapid segmentation of 9 key temporal bone structures in a data set of 325 clinical CTs. The data set was manually annotated by a specialist to serve as the ground truth. The data set was randomly split into training (n = 260) and testing (n = 65) sets. The model's performance was objectively assessed through external validation on the test set using metrics including Dice, Balanced accuracy, Hausdorff distances, and processing time. The model achieved an average Dice coefficient of 0.87 for all structures, an average balanced accuracy of 0.94, an average Hausdorff distance of 0.79 mm, and an average processing time of 9.1 seconds per CT. The present DL model for the automated simultaneous segmentation of multiple structures within the temporal bone from CTs achieved high accuracy according to currently commonly employed objective analysis. The results demonstrate the potential of the method to improve preoperative evaluation and intraoperative guidance in otologic surgery.

A convolutional neural network-based method for the generation of super-resolution 3D models from clinical CT images

Background and objectiveThe accurate evaluation of bone mechanical properties is essential for predicting fracture risk based on clinical computed tomography (CT) images. However, blurring and noise in clinical CT images can compromise the accuracy of these predictions, leading to incorrect diagnoses. Although previous studies have explored enhancing trabecular bone CT images to super-resolution (SR), none of these studies have examined the possibility of using clinical CT images from different instruments, typically of lower resolution, as a basis for analysis. Additionally, previous studies rely on 2D SR images, which may not be sufficient for accurate mechanical property evaluation, due to the complex nature of the 3D trabecular bone structures. The objective of this study was to address these limitations. MethodsA workflow was developed that utilizes convolutional neural networks to generate SR 3D models across different clinical CT instruments. The morphological and finite-element-derived mechanical properties of these SR models were compared with ground truth models obtained from micro-CT scans. ResultsA significant improvement in analysis accuracy was demonstrated, where the new SR models increased the accuracy by up to 700 % compared with the low-resolution data, i.e. clinical CT images. Additionally, we found that the mixture of different CT image datasets may improve the SR model performance. ConclusionsSR images, generated by convolutional neural networks, outperformed clinical CT images in the determination of morphological and mechanical properties. The developed workflow could be implemented for fracture risk prediction, potentially leading to improved diagnoses and subsequent clinical decision making.

Real-time germanium-doped optical fibers for clinical computed tomography dosimetry

In studies focusing on the practice of clinical computed tomography (CT) a particular concern has been that of delivering unnecessary dose to patients and with it the added risk for radiation-induced cancers. For a CT system, present work has investigated the dosimetric characteristics of a Ge-doped optical fiber real-time dosimetry system (model LS-2000, Lumisyns), measuring beam quality, exposure linearity and exposure duration accuracy. For comparison, the study has used a RaySafe X2 test device (Unfors RaySafe) and a Black Piranha (RTI) QA meter. Irradiations were made using a Somatom Definition Flash dual-source 128-slice CT scanner (Siemens Healthcare). The LS-2000, RaySafe X2, and Black Piranha sensors were exposed to beams energized at accelerating potentials ranging from 80 to 140 kVp, with tube current-time products from 50 to 300 mA, and exposure durations from 750 to 2000 ms. Constant signal responses from the LS-2000 have been obtained, measured in respect of beam quality, exposure linearity, and time accuracy. In respect of accuracy of beam quality, the RaySafe X2 and Black Piranha maximum deviations were 1% and 5%, respectively; the coefficients of linearity of RaySafe, Black Piranha and LS-2000 are 1%, 0.4% and 1.4%, respectively; the maximum deviation for the time accuracy test for the three sensors are 0.5%, 2% and 1.3%, respectively. Evaluation of the LS-2000 real-time dosimetry system with a Ge-doped optical fiber scintillator points to potential for its use in clinical CT dosimetry.

High resolution propagation-based lung imaging at clinically relevant X-ray dose levels

Absorption-based clinical computed tomography (CT) is the current imaging method of choice in the diagnosis of lung diseases. Many pulmonary diseases are affecting microscopic structures of the lung, such as terminal bronchi, alveolar spaces, sublobular blood vessels or the pulmonary interstitial tissue. As spatial resolution in CT is limited by the clinically acceptable applied X-ray dose, a comprehensive diagnosis of conditions such as interstitial lung disease, idiopathic pulmonary fibrosis or the characterization of small pulmonary nodules is limited and may require additional validation by invasive lung biopsies. Propagation-based imaging (PBI) is a phase sensitive X-ray imaging technique capable of reaching high spatial resolutions at relatively low applied radiation dose levels. In this publication, we present technical refinements of PBI for the characterization of different artificial lung pathologies, mimicking clinically relevant patterns in ventilated fresh porcine lungs in a human-scale chest phantom. The combination of a very large propagation distance of 10.7 m and a photon counting detector with 100,upmu hbox {m} pixel size enabled high resolution PBI CT with significantly improved dose efficiency, measured by thermoluminescence detectors. Image quality was directly compared with state-of-the-art clinical CT. PBI with increased propagation distance was found to provide improved image quality at the same or even lower X-ray dose levels than clinical CT. By combining PBI with iodine k-edge subtraction imaging we further demonstrate that, the high quality of the calculated iodine concentration maps might be a potential tool for the analysis of lung perfusion in great detail. Our results indicate PBI to be of great value for accurate diagnosis of lung disease in patients as it allows to depict pathological lesions non-invasively at high resolution in 3D. This will especially benefit patients at high risk of complications from invasive lung biopsies such as in the setting of suspected idiopathic pulmonary fibrosis (IPF).

Evaluation of cortical bone density using clinical computed tomography images: Detection of cortical porosity areas or transitional zones in human femoral diaphyses.

Age-related changes in human trabecular bone and cortical bone are known to vary. Although the porosity of cortical bone has been suggested to increase the risk of bone fracture, most of the currently available instruments for osteoporosis testing target trabecular bone. In this study, we evaluated cortical bone density using clinical computed tomography (CT) and compared the reliability of the cortical bone density index (CDI) with that of a polished male femoral bone from the same region. CDI images revealed that the porous area of cortical bone was extended in low CDI values. Moreover, this method was used to semi-quantitatively evaluate the cortical bones of the diaphysis of male femur specimens (n=46). We found that there was a significant relationship (r=0.70, p < 0.01) between the value of the cortical index (the ratio of cortical bone area to the cross-sectional area of the femoral diaphysis) and the average of CDI in the low signal area. Our findings suggest that the smaller the cortical bone occupancy, the more areas of consequential bone density loss were present. This may be the first step toward using clinical CT to assess cortical bone density.

Analysis of Three-Dimensional Bone Microarchitecture of the Axis Exposes Pronounced Regional Heterogeneity Associated with Clinical Fracture Patterns

The odontoid process (dens) of the second cervical vertebra (axis) is prone to fracture. While the importance of its skeletal integrity has been previously noted, representative three-dimensional microarchitecture analyses in humans are not available. This study aimed to determine the bone microarchitecture of the axis using high-resolution quantitative computed tomography (HR-pQCT) and to derive clinical implications for the occurrence and treatment of axis fractures. For initial clinical reference, the apparent density of the axis was determined based on clinical computed tomography (CT) images in patients without and with fractures of the axis. Subsequently, 28 human axes (female 50%) obtained at autopsy were analyzed by HR-pQCT. Analyses were performed in three different regions corresponding to zones I (tip of dens), II (base of dens), and III (corpus axis) of the Anderson and D’Alonzo classification. Lower apparent densities based on clinical CT data were detected in zone II and III compared to zone I in both the group without and with fracture. In the autopsy specimens, cortical thickness and bone volume fraction decreased continuously from zone I to zone III. Trabecular and cortical tissue mineral density was lowest in zone III, with no differences between zones I and II. In conclusion, our clinical and high-resolution ex vivo imaging data highlight a marked regional heterogeneity of bone microarchitecture, with poor cortical and trabecular properties near the dens base. These results may partly explain why zones II and III are at high risk of fracture and osteosynthesis failure.

Statistical Morphology and Fragment Mapping of Complex Proximal Humeral Fractures

Background and Objectives: Proximal humerus fractures (PHFs) are common in the elderly, but the treatment results are often poor. A clear understanding of fracture morphology and distribution of cortical bone loss is important for improved surgical decision making, operative considerations, and new implant designs. The aim of this study was to develop a 3D segmentation fracture mapping technique to create a statistical description of the spatial pattern and cortical bone loss of complex PHFs. Materials and Methods: Fifty clinical computed tomography (CT) scans of complex PHFs and their contralateral intact shoulders were collected. In-house software was developed for semi-automated segmentation and fracture line detection and was combined with manual fracture reduction to the contralateral template in a commercial software. A statistical mean model of these cases was built and used to describe probability maps of the fracture lines and cortical fragments. Results: The fracture lines predominantly passed through the surgical neck and between the tuberosities and tendon insertions. The superior aspects of the tuberosities were constant fragments where comminution was less likely. Some fracture lines passed through the bicipital sulcus, but predominantly at its edges and curving around the tuberosities proximally and distally. Conclusions: A comprehensive and systematic approach was developed for processing clinical CT images of complex fractures into fracture morphology and fragment probability maps and applied on PHFs. This information creates an important basis for better understanding of fracture morphology that could be utilized in future studies for surgical training and implant design.

Development and CT image-domain validation of a computational lung lesion model for use in virtual imaging trials.

Computational abnormalities (e.g., lesion models) for use in medical imaging simulation studies are frequently generated using data collected from clinical images. Although this approach allows for highly-customizable lesion detectability studies on clinical computed tomography (CT) data, the ground-truth lesion models produced with this method do not provide a sufficiently realistic lesion morphology for use with current anthropomorphic simulation studies. This work is intended to demonstrate that the new anatomically-informed lesion model presented here is not inferior to the previous lesion model under CT imaging, and can therefore provide a more biologically-informed model for use with simulated CT imaging studies. The lesion model was simulated initially from a seed cell with 10μm diameter placed in an anatomical location within segmented lung CT and was allowed to reproduce locally within the available solid angle in discrete time-intervals (corresponding to synchronous cell cycles) up to a size of ∼200μm in diameter. Daughter cells of generation G were allowed also to reproduce on the next available time-step given sufficient space. At lesion sizes beyond 200μm in diameter, the health of subregions of cells were tracked with a Markov chain technique, indicating which regions were likely to continue growing, which were likely stable, and which were likely to develop necrosis given their proximity to anatomical features and other lesion cells. For lesion sizes beyond 500μm, the lesion was represented with three nested, triangulated surfaces (corresponding to proliferating, dormant, and necrotic regions), indicating how discrete volumes of the lesion were behaving at a particular time. Lesions were then assigned smoothly-varying material properties based on their cellular level health in each region, resulting in a multi-material lesion model. The lesions produced with this model were then voxelized and placed into lung CT images for comparison with both prior work and clinical data. This model was subject to an observer study in which cardiothoracic imaging radiologists assessed the realism of both clinical and synthetic lesions in CT images. The useable outputs of this work were voxel- or surface-based, validated, computational lesions, at a scale clearly visible on clinical CT (3-4cm). Analysis of the observer study results indicated that the computationally-generated lesions were indistinguishable from clinical lesions (AUC=0.49, 95% CI=[0.36, 0.61]) and non-inferior to an earlier image-based lesion model-indicating the advantage of the model for use in both hybrid CT images and in simulated CT imaging of the lungs. Results indicated the non-inferiority of this model as compared to previous methods, indicating the utility of the model for use in both hybrid CT images and in simulated CT imaging.

Quantifying the Human Subchondral Trabecular Bone Microstructure in Osteoarthritis with Clinical CT.

Osteoarthritis (OA) is characterized by critical alterations of the subchondral bone microstructure, besides the well‐known cartilaginous changes. Clinical computed tomography (CT) detection of quantitative 3D microstructural subchondral bone parameters is applied to monitor changes of subchondral bone structure in different stages of human OA and is compared with micro‐CT, the gold standard. Determination by clinical CT (287 µm resolution) of key microstructural parameters in tibial plateaus with mild‐to‐moderate and severe OA reveals strong correlations to micro‐CT (35 µm), high inter‐ and intraobserver reliability, and small relative differences. In vivo, normal, mild‐to‐moderate, and severe OA are compared with clinical CT (331 µm). All approaches detect characteristic expanded trabecular structure in severe OA and fundamental microstructural correlations with clinical OA stage. Multivariate analyses at various in vivo and ex vivo imaging resolutions always reliably separate mild‐to‐moderate from severe OA (except mild‐to‐moderate OA from normal), revealing a striking similarity between 287 µm clinical and 35 µm micro‐CT. Thus, accurate structural measurements using clinical CT with a resolution near the trabecular dimensions are possible. Clinical CT offers an opportunity to quantitatively monitor subchondral bone microstructure in clinical and experimental settings as an advanced tool of investigating OA and other diseases affecting bone architecture.

SR-CycleGAN: super-resolution of clinical CT to micro-CT level with multi-modality super-resolution loss.

Purpose: We propose a super-resolution (SR) method, named SR-CycleGAN, for SR of clinical computed tomography (CT) images to the micro-focus x-ray CT CT ( ) level. Due to the resolution limitations of clinical CT (about ), it is challenging to obtain enough pathological information. On the other hand, scanning allows the imaging of lung specimens with significantly higher resolution (about or higher), which allows us to obtain and analyze detailed anatomical information. As a way to obtain detailed information such as cancer invasion and bronchioles from preoperative clinical CT images of lung cancer patients, the SR of clinical CT images to the level is desired. Approach: Typical SR methods require aligned pairs of low-resolution (LR) and high-resolution images for training, but it is infeasible to obtain precisely aligned paired clinical CT and images. To solve this problem, we propose an unpaired SR approach that can perform SR on clinical CT to the level. We modify a conventional image-to-image translation network named CycleGAN to an inter-modality translation network named SR-CycleGAN. The modifications consist of three parts: (1)an innovative loss function named multi-modality super-resolution loss, (2)optimized SR network structures for enlarging the input LR image to -times by width and height to obtain the SR output, and (3)sub-pixel shuffling layers for reducing computing time. Results: Experimental results demonstrated that our method successfully performed SR of lung clinical CT images. SSIM and PSNR scores of our method were 0.54 and 17.71, higher than the conventional CycleGAN's scores of 0.05 and 13.64, respectively. Conclusions: The proposed SR-CycleGAN is usable for the SR of a lung clinical CT into scale, while conventional CycleGAN output images with low qualitative and quantitative values. More lung micro-anatomy information could be observed to aid diagnosis, such as the shape of bronchioles walls.

Comparison of clinical-computed tomography model with 2D and 3D radiomics models to predict occult peritoneal metastases in advanced gastric cancer.

To compare the ability of a clinical-computed tomography (CT) model vs. 2D and 3D radiomics models for predicting occult peritoneal metastasis (PM) in patients with advanced gastric cancer (AGC). In this retrospective study, we included 49 patients with occult PM and 49 control patients (without PM) who underwent preoperative CT and subsequent surgery between January 2016 and December 2018. Clinical information and CT semantic features were collected, and CT radiomics features were extracted. A predictive clinical-CT model was created using multivariate logistic regression. The least absolute shrinkage and selection operator algorithm and logistic regression were used for constructing 2D and 3D radiomics models.These models were validated with an external cohort (n = 30). Receiver operating characteristics curve with area under the curve (AUC), sensitivity, and specificity were used to evaluate predictive performance. Tumor size, mild ascites, and serum CA125 were independent factors predictive of occult PM. The clinical-CT model of these independent factors showed better diagnostic performance than 2D and 3D radiomics models. In the external validation cohort, the AUCs of different models were as follows-clinical-CT model: 0.853 (sensitivity, 66.7%; specificity, 93.3%); 2D radiomics model: 0.622 (sensitivity, 80.0%; specificity, 46.7%); and 3D radiomics model: 0.676 (sensitivity, 60.0%; specificity, 86.0%). The clinical-CT model nomogram showed good clinical predictive efficiency to assess occult PM. The clinical-CT model was better than the radiomics models in predicting occult PM in AGC.

Recent advances in airway imaging using micro-computed tomography and computed tomography for chronic obstructive pulmonary disease.

Chronic obstructive pulmonary disease (COPD) is a complex lung disease characterized by a combination of airway disease and emphysema. Emphysema is classified as centrilobular emphysema (CLE), paraseptal emphysema (PSE), or panlobular emphysema (PLE), and airway disease extends from the respiratory, terminal, and preterminal bronchioles to the central segmental airways. Although clinical computed tomography (CT) cannot be used to visualize the small airways, micro-CT has shown that terminal bronchiole disease is more severe in CLE than in PSE and PLE, and micro-CT findings suggest that the loss and luminal narrowing of terminal bronchioles is an early pathological change in CLE. Furthermore, the introduction of ultra-high-resolution CT has enabled direct evaluation of the proximal small (1 to 2-mm diameter) airways, and new CT analytical methods have enabled estimation of small airway disease and prediction of future COPD onset and lung function decline in smokers with and without COPD. This review discusses the literature on micro-CT and the technical advancements in clinical CT analysis for COPD. Hopefully, novel micro-CT findings will improve our understanding of the distinct pathogeneses of the emphysema subtypes to enable exploration of new therapeutic targets, and sophisticated CT imaging methods will be integrated into clinical practice to achieve more personalized management.

Fully Automated Measurement of Cochlear Duct Length From Clinical Temporal Bone Computed Tomography.

To present and validate a novel fully automated method to measure cochlear dimensions, including cochlear duct length (CDL). Cross-sectional study. The computational method combined 1) a deep learning (DL) algorithm to segment the cochlea and otic capsule and 2) geometric analysis to measure anti-modiolar distances from the round window to the apex. The algorithm was trained using 165 manually segmented clinical computed tomography (CT). A Testing group of 159 CTs were then measured for cochlear diameter and width (A- and B-values) and CDL using the automated system and compared against manual measurements. The results were also compared with existing approaches and historical data. In addition, pre- and post-implantation scans from 27 cochlear implant recipients were studied to compare predicted versus actual array insertion depth. Measurements were successfully obtained in 98.1% of scans. The mean CDL to 900° was 35.52 mm (SD, 2.06; range, [30.91-40.50]), the mean A-value was 8.88 mm (0.47; [7.67-10.49]), and mean B-value was 6.38 mm (0.42; [5.16-7.38]). The R2 fit of the automated to manual measurements was 0.87 for A-value, 0.70 for B-value, and 0.71 for CDL. For anti-modiolar arrays, the distance between the imaged and predicted array tip location was 0.57 mm (1.25; [0.13-5.28]). Our method provides a fully automated means of cochlear analysis from clinical CTs. The distribution of CDL, dimensions, and cochlear quadrant lengths is similar to those from historical data. This approach requires no radiographic experience and is free from user-related variation. 3 Laryngoscope, 132:449-458, 2022.

A new tool for medical image processing in biological anthropology and anatomy research

The use of clinical computed tomography (CT) scans as sources for data in biological anthropology and anatomy research is becoming increasingly common. While it is best to access the original high-resolution scans, the reformatted stacks of low-resolution multiplanar reconstructions are the most commonly available to researchers. Volumetric skeletal models reconstructed from these low-resolution scans can lead to distorted measurements and/or cause problems in subsequent analyses. This study investigates the utility of NiftyMIC, a python-based application, in registering and reconstructing high-resolution volume from multiple stacks of low-resolution scans. We do so by comparing the skeletal measurements taken on the resulting 3D long bone models to those reconstructed using a “classic” semi-automatic segmentation protocol from original low-resolution scans. CT scans of 33 individuals aged 0 to 16 years were collected from National Taiwan University Hospital. Two sets of linear measurements, one on the original low-resolution volume reconstructions and one on the registered high-resolution volume reconstructions, were collected by two independent investigators. A total of 197 linear measurements (80 lengths and 117 breadths) was collected. Percent differences were used to assess the accuracy of measurements taken on bones using each reconstruction. Technique while technical error of measurement (TEM) and relative technical error of measurement (%TEM) were used to test the reliability of the registration and reconstruction processes and observer errors. The results showed low and comparable percent differences for original reconstructions (0.688 to 3.801%) and the registered reconstructions (0.202 to 3.47%). Inter-modality TEM and %TEM were 0.089mm and 0.113%, respectively, showing high reliability of the registering technique. Intra- and inter-observer errors were comparable on both types of reconstructions and lower than 1mm and 1% for all TEMs and %TEMs. The registered high-resolution volume reconstructions provided much smoother surfaces and a clean separation of fusing elements (epiphyses and diaphyses) (Figure 1). This study demonstrates the utility of a new image processing tool in registering and reconstructing high-resolution volumes from low-resolution clinical CT scans. The application is especially useful in studies of children and juvenile individuals where separation of fusing skeletal elements is required, and in shape analyses, where clean and precise surfaces are beneficial.

A Benchmark for automatic noise measurement in clinical computed tomography.

Assessment of image quality directly in clinical image data is an important quality control objective as phantom-based testing does not fully represent image quality across patient variation. Computer algorithms for automatically measuring noise in clinical computed tomography (CT) images have been introduced, but the accuracy of these algorithms is unclear. This work benchmarks the accuracy of the global noise (GN) algorithm for automatic noise measurement in contrast-enhanced abdomen CT exams in comparison to precise reference noise measurements. The GN algorithm was further optimized compared to the previous report in the literature. Reference values of noise were established in a public image dataset of 82 contrast-enhanced abdomen CT exams. The reference noise values were obtained by manual regions-of-interest measurements of pixel standard deviation in the liver parenchyma according to an instruction protocol. Noise measurements taken by six observers were averaged together to improve reference noise statistical precision. The GN algorithm was used to automatically measure noise in each image set. The accuracy of the GN algorithm was determined in terms of RMS error compared to reference noise. The GN algorithm was optimized by conducting 1000 trials with random algorithm parameter values. The trial with the lowest RMS error was used to select optimum algorithm parameters. The range of noise across CT image sets was 8.8-28.8 HU. Reference noise measurements were made with a precision of ±0.78HU (95% confidence interval). The RMS error of automatic noise measurement was 0.93HU (0.77-1.19HU 95% confidence interval). The automatic noise measurements were equally accurate across image sets of varying noise magnitude. Optimum GN algorithm parameter values were: a kernel size of 7 pixels, and soft tissue lower and upper thresholds of 0 and 170HU, respectively. The performance of automatic noise measurement was benchmarked in a large clinical CT dataset. The study provides a framework for thorough validation of automatic clinical image quality measurement methods. The GN algorithm was optimized and validated for automatic measurement of soft-tissue noise in abdomen CT exams.

Monitoring treatment effects in lung cancer-bearing mice: clinical CT and clinical MRI compared to micro-CT

BackgroundCompared to histology-based methods, imaging can reduce animal usage in preclinical studies. However, availability of dedicated scanners is limited. We evaluated clinical computed tomography (CT) and magnetic resonance imaging (MRI) in comparison to dedicated CT (micro-CT) for assessing therapy effects in lung cancer-bearing mice.MethodsAnimals received cisplatin (n = 10), sham (n = 12), or no treatment (n = 9). All were examined via micro-CT, CT, and MRI before and after treatment. Semiautomated tumour burden (TB) calculation was performed. The Bland-Altman, receiver operating characteristic (ROC), and Spearman statistics were used.ResultsAll modalities always allowed localising and measuring TB. At all modalities, mice treated with cisplatin showed a TB reduction (p ≤ 0.012) while sham-treated and untreated individuals presented tumour growth (p < 0.001). Mean relative difference (limits of agreement) between TB on micro-CT and clinical scanners was 24.7% (21.7–27.7%) for CT and 2.9% (−4.0–9.8%) for MRI. Relative TB changes before/after treatment were not different between micro-CT and CT (p = 0.074) or MRI (p = 0.241). Mice with cisplatin treatment were discriminated from those with sham or no treatment at all modalities (p ≤ 0.001). Using micro-CT as reference standard, ROC areas under the curves were 0.988–1.000 for CT and 0.946–0.957 for MRI. TB changes were highly correlated across modalities (r ≥ 0.900, p < 0.001).ConclusionsClinical CT and MRI are suitable for treatment response evaluation in lung cancer-bearing mice. When dedicated scanners are unavailable, they should be preferred to improve animal welfare.

Validation of algorithmic CT image quality metrics with preferences of radiologists

Automated assessment of perceptual image quality on clinical Computed Tomography (CT) data by computer algorithms has the potential to greatly facilitate data-driven monitoring and optimization of CT image acquisition protocols. The application of these techniques in clinical operation requires the knowledge of how the output of the computer algorithms corresponds to clinical expectations. This study addressed the need to validate algorithmic image quality measurements on clinical CT images with preferences of radiologists and determine the clinically acceptable range of algorithmic measurements for abdominal CT examinations. Algorithmic measurements of image quality metrics (organ HU, noise magnitude, and clarity) were performed on a clinical CT image dataset with supplemental measures of noise power spectrum from phantom images using techniques developed previously. The algorithmic measurements were compared to clinical expectations of image quality in an observer study with seven radiologists. Sets of CT liver images were selected from the dataset where images in the same set varied in terms of one metric at a time. These sets of images were shown via a web interface to one observer at a time. First, the observer rank ordered the CT images in a set according to his/her preference for the varying metric. The observer then selected his/her preferred acceptable range of the metric within the ranked images. The agreement between algorithmic and observer rankings of image quality were investigated and the clinically acceptable image quality in terms of algorithmic measurements were determined. The overall rank-order agreements between algorithmic and observer assessments were 0.90, 0.98, and 1.00 for noise magnitude, liver parenchyma HU, and clarity, respectively. The results indicate a strong agreement between the algorithmic and observer assessments of image quality. Clinically acceptable thresholds (median) of algorithmic metric values were (17.8, 32.6) HU for noise magnitude, (92.1, 131.9) for liver parenchyma HU, and (0.47, 0.52) for clarity. The observer study results indicated that these algorithms can robustly assess the perceptual quality of clinical CT images in an automated fashion. Clinically acceptable ranges of algorithmic measurements were determined. The correspondence of these image quality assessment algorithms to clinical expectations paves the way toward establishing diagnostic reference levels in terms of clinically acceptable perceptual image quality and data-driven optimization of CT image acquisition protocols.

Correlation of Algorithmic and Visual Assessment of Lesion Detection in Clinical Images

Clinically-relevant quantitative measures of task-based image quality play key roles in effective optimization of medical imaging systems. Conventional phantom-based measures do not adequately reflect the real-world image quality of clinical Computed Tomography (CT) series which is most relevant for diagnostic decision-making. The assessment of detectability index which incorporates measurements of essential image quality metrics on patient CT images can overcome this limitation. Our current investigation extends and validates the technique on standard-of-care clinical cases. We obtained a clinical CT image dataset from an Institutional Review Board-approved prospective study on colorectal adenocarcinoma patients for detecting hepatic metastasis. For this study, both perceptual image quality and lesion detection performance of same-patient CT image series with standard and low dose acquisitions in the same breath hold and four processing algorithms applied to each acquisition were assessed and ranked by expert radiologists. The clinical CT image dataset was processed using the previously validated method to estimate a detectability index for each known lesion size in the size distribution of hepatic lesions relevant for the imaging task and for each slice of a CT series. We then combined these lesion-size-specific and slice-specific detectability indexes with the size distribution of hepatic lesions relevant for the imaging task to compute an effective detectability index for a clinical CT imaging condition of a patient. The assessed effective detectability indexes were used to rank task-based image quality of different imaging conditions on the same patient for all patients. We compared the assessments to those by expert radiologists in the prospective study in terms of rank order agreement between the rankings of algorithmic and visual assessment of lesion detection and perceptual quality. Our investigation indicated that algorithmic assessment of lesion detection and perceptual quality can predict observer assessment for detecting hepatic metastasis. The algorithmic and visual assessment of lesion detection and perceptual quality are strongly correlated using both the Kendall's Tau and Spearman's Rho methods (perfect agreement has value 1): for assessment of lesion detection, 95% of the patients have rank correlation coefficients values exceeding 0.87 and 0.94, respectively, and for assessment of perceptual quality, 0.85 and 0.94, respectively. This study used algorithmic detectability index to assess task-based image equality for detecting hepatic lesions and validated it against observer rankings on standard-of-care clinical CT cases. Our study indicates that detectability index provides a robust reflection of overall image quality for detecting hepatic lesions under clinical CT imaging conditions. This demonstrates the concept of utilizing the measure to quantitatively assess the quality of the information content that different imaging conditions can provide for the same clinical imaging task, which enables targeted optimization of clinical CT systems to minimize clinical and patient risks.

Skeletal muscle cutoff values for sarcopenia diagnosis using T10 to L5 measurements in a healthy US population

Measurements of skeletal muscle cross-sectional area, index, and radiation attenuation utilizing clinical computed tomography (CT) scans are used in assessments of sarcopenia, the loss of skeletal muscle mass and function associated with aging. To classify individuals as sarcopenic, sex-specific cutoffs for ‘low’ values are used. Conventionally, cutoffs for skeletal muscle measurements at the level of the third lumbar (L3) vertebra are used, however L3 is not included in several clinical CT protocols. Non-contrast-enhanced CT scans from healthy kidney donor candidates (age 18–40) at Michigan Medicine were utilized. Skeletal muscle area (SMA), index (SMI), and mean attenuation (SMRA) were measured at each vertebral level between the tenth thoracic (T10) and the fifth lumbar (L5) vertebra. Sex-specific means, standard deviations (s.d.), and sarcopenia cutoffs (mean-2 s.d.) at each vertebral level were computed. Associations between vertebral levels were assessed using Pearson correlations and Tukey’s difference test. Classification agreement between different vertebral level cutoffs was assessed using overall accuracy, specificity, and sensitivity. SMA, SMI, and SMRA L3 cutoffs for sarcopenia were 92.2 cm2, 34.4 cm2/m2, and 34.3 HU in females, and 144.3 cm2, 45.4 cm2/m2, and 38.5 HU in males, consistent with previously reported cutoffs. Correlations between all level pairs were statistically significant and high, ranging from 0.65 to 0.95 (SMA), 0.64 to 0.95 (SMI), and 0.63 to 0.95 (SMRA). SMA peaks at L3, supporting its use as the primary site for CT sarcopenia measurements. However, when L3 is not available alternative levels (in order of preference) are L2, L4, L5, L1, T12, T11, and T10. Healthy reference values reported here enable sarcopenia assessment and sex-specific standardization of SMA, SMI, and SMRA in clinical populations, including those whose CT protocols do not include L3.